HapticPrint: Designing Feel Aesthetics for 3D Printing

Cesar Torres Tim Campbell Neil Kumar Eric PaulosElectrical Engineering and Computer Sciences

University of California Berkeley{cearto, timcam, n.kumar, paulos}@berkeley.edu

Figure 1. Example objects created with HapticPrint. External and Internal tools allow users to easily add tactility, flexibility, and weight to objects.

ABSTRACTDigital fabrication has enabled massive creativity in hobbyistcommunities and professional product design. These emerg-ing technologies excel at realizing an arbitrary shape or form;however these objects are often rigid and lack the feel desiredby designers. We aim to enable physical haptic design in pas-sive 3D printed objects. This paper identifies two core areasfor extending physical design into digital fabrication: design-ing the external and internal haptic characteristics of an ob-ject. We present HapticPrint as a pair of design tools to easilymodify the feel of a 3D model. Our external tool maps tex-tures and UI elements onto arbitrary shapes, and our internaltool modifies the internal geometry of models for novel com-pliance and weight characteristics. We demonstrate the valueof HapticPrint with a range of applications that expand theaesthetics of feel, usability, and interactivity in 3D artifacts.

Author Keywordsdigital fabrication; haptics; design; texture; deformation

ACM Classification KeywordsD.2.2 Design Tools and Techniques: User interfaces (H.5.2,H.1.2, I.3.6)

INTRODUCTIONIn current digital fabrication practices, 3D modeling has oftenbeen seen as a medium concerned with the look of an artifact.Yet designing the feel of an artifact is highly critical to the us-ability, accessibility, and perception of physical artifacts [15].

Permission to make digital or hard copies of all or part of this work for personal orclassroom use is granted without fee provided that copies are not made or distributedfor profit or commercial advantage and that copies bear this notice and the full citationon the first page. Copyrights for components of this work owned by others than theauthor(s) must be honored. Abstracting with credit is permitted. To copy otherwise, orrepublish, to post on servers or to redistribute to lists, requires prior specific permissionand/or a fee. Request permissions from Permissions@acm.org.UIST 2015, November 8–11, 2015, Charlotte, NC, USA.Copyright is held by the owner/author(s). Publication rights licensed to ACM.ACM 978-1-4503-3779-3/15/11 ...$15.00.DOI: http://dx.doi.org/10.1145/2807442.2807492

The feel aesthetic, or the choice of haptic characteristics givento an object, contributes to both the functional and aestheticvalue of an artifact such as adding a knurl1 texture to a sim-ple knob. Albeit subtle, these cues are a part of the design ofeveryday objects.

Primarily influenced by form-first design tools and limitedaccess to multi-material 3D printers, today’s artifacts are pro-duced as rigid, smooth plastic parts. We see this as an op-portunity to enable the design of haptic characteristics andexpand the feel aesthetic of printed artifacts using current 3Dprinting techniques. While several methods have been devel-oped for expanding the range of haptic properties in digitallyfabricated artifacts [1, 4, 19, 21, 22], synthesizing these hap-tic characteristics in a design tool remains underdeveloped.

We introduce HapticPrint, a pair of design tools that enableusers to easily design the feel aesthetics of an object usingFused Filament Fabrication (FFF). In both tools, we distill the3D design problem into a 2D design task for users. Our ex-ternal tool overlays textures derived from 2D raster graphicsto generate tactile models on a variety of surfaces. Our inter-nal tool controls a 3D model’s stiffness and mass distributionby partitioning it into “chambers” which can then be ascribedspecial haptic infills. Features of infill and texture patternswere parametrized to enable users to control specific cues ofhaptic exploration (Figure 2). Lastly, to aid users with hapticselection and facilitate an iterative design practice, we createdan online library of printable reference designs, or palettes.

We motivate HapticPrint with a review of related work inhaptic perception, digital fabrication, and the physical designspace. We then outline the implementation of the two Hap-ticPrint tools and evaluate each Feel Aesthetic. Finally ap-plications with HapticPrint show how the synthesis of hapticdesign produces novel artifacts and interactions.

1Knurl is a pattern of angled cut lines given to machined parts foruser grips.

1

RELATED WORKOur work aims to connect research on feel aesthetics and pro-totyping tools for more expressive and usable objects. We re-viewed literature on haptic perception, prototyping in design,and 3D printing tools and techniques.

Prototyping FeelHoude & Hill’s work on iterative design identifies look &feel as critical properties of successful prototypes [11], re-ferring to both the interactive feel of virtual interfaces and thephysical feel of tangible prototypes. Despite the importanceof haptic characteristics in an object’s design, haptic designremains an underdeveloped area. Instead, design languagefor feel largely borrows from visual design [29]. In seminalwork on Haptic Exploration, Lederman & Klatzky showedthat users identify objects by six exploratory procedures: lat-eral motion (texture), pressure (hardness), static contact (tem-perature), unsupported holding (weight), enclosure (volume),and contour following (global shape) [16]. We use this hapticexploration space to guide the design of HapticPrint (Figure2). While 3D printing for prototyping largely captures theglobal shape of an artifact, we seek to develop techniques foraddressing a larger breadth of haptic exploration language, orfeel, of a 3D printed artifact.

Material PrototypingIn order to expand the range of producible 3D printed arti-facts, recent research explores controlling or simulating ma-terial properties other than thermoplastics. Bickel et. al. useda multi-material printer to synthesize target material behav-iors by layering two or more base materials [1]. This ma-terial “dithering” is used to produce materials with hybridproperties. This work led to the development of OpenFab,a shader-like specification language for the multi-materialprinting pipeline [32]. Other fabrication techniques work di-rectly with the desired materials, including wool [12] or car-bon fiber2. While Bickel et al. approaches haptic design froma goal-driven framework, we focus on the iterative and re-flective design cycle used by many product designers. Moresuccinctly, how does a designer choose the right feel?

Surface Design: On the Skin3D modeling remains a barrier to many amateur designers,let alone modeling surface textures. Simplified methods suchas 2.1D models (i.e., stacked planar slices) have the poten-tial to increase participation in 3D design [24]. Similarly,2.5D models have been constructed from Shape from Shad-ing (SFS) or sensor-based depth-maps to create tactile paint-ings [5], graphs [2], and maps [8]. More recently, these tech-niques have been used to create physical visualizations ofdata [28]. More sophisticated tools such as ZBrush3 unwrapa 3D mesh (UV mapping) on a 2D canvas allowing users tospecify texture through brush interactions.

Surface textures can also be sampled directly from the en-vironment. Using depth sensors Li et al. reconstructedfull body portraits from noisy environments [17]. Further-more, post-capture refinement algorithms have been used to2http://www.markforged.com3http://www.pixologic.com/

successfully capture and abstract textures such as hair [4].However, capturing detailed textures requires high-resolutionmeshes and introduce a large computational overhead for in-teractive applications. By generating textures from imported2D height maps, users can manipulate textures in a real-timeenvironment. Displacement mapping, the equivalent tech-nique in computer graphics, uses a GPU vertex shader to cal-culate a vertex’s 2D position on a screen to later result in a 2Draster graphic; our technique leverages HTML5 WebWorkersto calculate a vertex’s 3D displacement and construct a 3Dmodel at interactive speeds.

Internal Design: Under the SkinMany new design opportunities arise from modifying the in-ternal structure of objects. For instance reducing the needfor an internal support structure via skin-frames [33] or wire-forms [21] can speed up the prototyping process and decreaseprint time by tenfold. Stava et al. demonstrated that a com-bination of thickening, hollowing, and supporting struts in-creases the structural integrity of a print [27]. Similarly, ahollow honeycomb interior was shown to add to an object’sdurability [19]. While these techniques generally maintainthe shape and form of an object, we contribute techniques forprototyping the feel of a model.

In digital fabrication design, Prevost et al. [23] modify the in-ternal voxel geometry of a 3D model in order to redistributean artifact’s center-of-gravity to achieve balance. More dy-namic behaviors have been explored using layered soft com-posite materials and pneumatic actuation [37].Likewise otherinternal modifications have been shown to enable outputmechanisms with optical light pipes [35], identification struc-tures [36], or internal media-containing tubes [25]. Whileothers have demonstrated it in principle, we look at internalstructure design under the lens of interaction design. We ex-plore how weight and compliance might be used in designs asas a mode of interaction and present a method to both redirectthe mass and compliance of different areas using standard 3Dprinting techniques.

THE PHYSICAL HAPTIC DESIGN SPACEPhysical design generally refers to giving shape and externalform to an artifact, yet design literature identifies several de-sign parameters like weight, texture, and resilience that areimportant to the feel, or haptic design, of an object [15]. Ac-cording to ISO standards, haptics are divided into tactile orcutaneous touch referring to mechanical stimulation of theskin, and kinaesthetics referring to that sensation of move-ment, force, and position of the body with respect to an ob-ject [31]. Loomis and Lederman further distinguish hapticsbased on a) active/passive touch i.e. the presence or absenceof motor control, and b) the afferent inputs used i.e., cuta-neous, kinesthetic, and haptic4 [14, 18]. In this paper, werestrict our exploration to active touch. In this section, wedescribe the haptic design elements used in HapticPrint andhow they are utilized to expand the physical haptic designspace.

4Haptic in this context is used to refer to combined inputs from thecutaneous and kinesthetic touch

2

Cutaneous TactilityTactility refers to the sense of touch derived from skin’sphysical contact with the environment [31]. Tactile percep-tion is limited by spatial sensitivity (≈ 1.5–2.5 mm peri-ods, dependent on force exerted on the skin) and temporalsensitivity (≈ 20, 40, 250 Hz bandpass mechanoreceptors);optimal information transfer can be achieved by matchingthe spatial and temporal display parameters to sensing lim-its [18]. Atomic perceptual models from psychophysics re-veal a feature space for tactile surfaces across three dimen-sions: roughness & smoothness, hardness & softness, andelasticity (springiness) [10]. These haptic feedback mecha-nisms are often used to communicate affordances i.e., cuesto a user that signal opportunities to perform an action [22].For example, a ridged texture around a camera lens conveyswhere a user should grip and twist for manual focus. We di-vide the space of tactile design under physical characteristicsand semiotic characteristics.

Physically, surface geometries and material properties largelycontribute to tactile sensing: for cutaneous interactions, hap-tic slip and torque feels has been linked to skin-stretch andslip-direction [26]. These surface geometries can be char-acterized by feature-size and the gap between features, orfeature-gap. Other derivable dimensions include fill-densityand volume-factor. The arrangement of the features withina texture also affects the user’s perception. Grids communi-cate order, honeycombs denote more natural formations, and“randomness” is closely associated with a digital aesthetic.Dynamic textures, or those that change over time, have beenexplored in materials like leather that develop a patina. Cer-tain textures need to retain a specific size and density in orderto convey its message (e.g., snake scales). We draw fromsemiotics in order to characterize these types of textures:

• indexical - references real world stimuli e.g. cloth;• iconic - abstracted stimuli e.g. studded scales, knurl;• symbolic - does not reference any real world stimuli, must

be learned e.g. braille, letters.

KinesthesiaKinesthetics is also referred to as proprioception and refers tosensory inputs from mechanoreceptors in the body’s muscles,tendons, and joints; however the ways these receptors mediateperception is less defined than cutaneous touch [3, 14].

In haptic design, kinesthetics are integral to haptic explo-ration; the ability of a body to exert forces and evaluate howan object alters normal body motion allows us to gain an un-derstanding of the kinesthetic properties of that object. Forinstance, a user applies pressure to a flexible object to under-stand how it should be held or attached. The visibility of theinner structure provides visual feedback of the mechanics ofthe object. For example, corrugation patterns in cardboardconvey the grain of flexibility, while the fine mesh of Sty-rofoam in foamboard conveys a softer more impressionableinteraction.

Other properties such as weight plays an important role in thetrustworthiness of an object [29]. For instance, a light drillis perceived as less trustworthy and powerful than a heavier

INTERNAL DESIGN TOOLEXTERNAL DESIGN TOOL

Unsupported HoldingPressureContour followingLateral motion

KINESTHETICSCUTANEOUS TOUCH

Figure 2. HapticPrint divides feel aesthetics into external and internaltools and in order to design haptic exploration [16] in physical design.Enclosure (not pictured) is supported by current modeling tools. Staticcontact is left for future work.

drill. The distribution of weight is also an indicator of stablestates - how an object sits or rests in the hand implies inter-action points and modes. For instance, many objects havea single base weight pattern where the weight distribution“grounds” the object, inviting a user to interact from the top ormerely implying a decorative “statue-esque” role. A weightdistribution could convey how interfaces should be held, suchas placing batteries in the lower cavity of a remote control toact as a cantilever for pressing buttons in the upper segment.Axial weight configurations encourage users to rotate objectsaround an axis. Polycentric centers-of-mass such as the loosepebbles in a rainstick encourage interaction by influencing theuser to switch between states.

We implemented specific subsets of this broad haptic designspace into our HapticPrint tools which are detailed in the nextsection.

DESIGN TOOL IMPLEMENTATIONWith HapticPrint, we implemented two physical design toolsfor adding feel aesthetics to models. Each tool addresseselements of the exploratory haptic space depicted in Figure2. The external tool overlays patterns on models allowingusers to specify tactile cues (lateral motion) or seams to fol-low (contour following). The internal tool allows a user tospecify compliant infill patterns (pressure) in specific regions,including an infill pattern for injecting weight post-print (un-supported holding). In this section, we detail the technicalimplementation of each tool and the process used to create alibrary of 25 feel textures, 5 structural infills, and 6 weightprofiles, each of which spans the feel design space. Figure 1shows shape primitives with example feel aesthetics.

EXTERNAL DESIGN TOOLThe primary goal of the external tool is to easily modify thesurface of a model with texture. In particular, the tool shouldbe capable of expressing a broad range of textures on a givenform. With this tool a designer can either import, create, orspecify an existing texture. While selection is fairly trivial ina digital design scenario, physical design is better facilitatedwith tangible references. As such, we see such a design toolthat supports capturing textures from the environment, de-signing textures virtually with physical perception feedback,and specifying unknown textures guided by similarity to otherknown textures.

3

BUMPSF50 G32 S4

feature_size gap_size smoothness_index

A B

D

C

Figure 3. The external design tool. (a) Library of textures, interactive patterns, and infill samples. (b) A height map generator patterning a “spike”feature. (c) A 2D height map is converted into a 2.5D tactile mesh using WebGL. (d) Example textures on swatch palettes.

Using three.js5 we created a web application to generatea 2.5D model from a raster graphic (Figure 3c). 2.5D refersto models that are 3-dimensional in appearance but only varyalong a single dimension; they can be represented by theheight field function of z = f(x, y) [24].

The external tool first converts a raster image into aheightmap based on the corresponding grayscale value ofeach pixel. Our tool then uses height displacement – a com-mon computer graphics technique – to modify points on themesh. The 2D raster graphic is UV mapped, or projected ontothe mesh, and used to bind the heightmap to mesh vertices.

To obtain high resolution height displacement, the heightmapneeds to be assigned to a high-resolution triangular mesh sur-face. For 2D rendering, this operation is offset to a vertexshader on the GPU; this unfortunately only calculates the 2Ddisplacement of vertex coordinate on the screen. In order toproduce a 3D mesh, we used a map-and-reduce routine withHTML5 WebWorkers to calculate vertex displacement. Forreference, operating on a mesh with 45K vertices, we reportthe following performance values: GPU vertex shader (5ms,2D graphic only), iterative (11-12s, mesh), and WebWorkermap/reduce (140ms, mesh).

In our implementation, we exposed the vertex-shader imageto the user and computed the full mesh displacement in thebackground. Height displacement occurs with respect to thedirection of the surface normal at each vertex. Finally, thetool stores the resulting mesh as an STL for 3D printing.

A clear advantage of a height displacement approach is thedecomposition of a 3D surface task into a 2D graphic task.Texture as a 2D input allows inspiration to come from ex-isting images rather than solely designing in a 3D modelingenvironment. Furthermore, many 2D graphic tools supporttools such as gradients and opacity, which can be used to cre-ate bumps, seams, and levels of elevation - a trait we exploitto enable tactile user interface (UI) design.

Importing, creating, and selecting a textureThe external tool takes textures in the form of a rastergraphic or Support Vector Graphic (SVG). Raster imageswere sourced from simple image searches as well as reposito-ries of bump maps6 in the graphics community. Parameteriz-able height maps were generated using a custom paper.jsapplication shown in Figure 3a.5http://threejs.org/6https://www.filterforge.com/filters/

A user can specify the individual feature element in a texture(i.e. a single cell of honeycomb), and the tool lets the userpattern and scale the texture based on feature size, featuregap, and arrangement.

The arrangements are a rectilinear grid, a honeycomb grid, ora honeycomb with added perlin noise for more natural forma-tions. For prototyping many textures, a designer can constructhaptic swatches as 70 mm x 70 mm 2.5D layers and placethem on a “swatch palette” similar to a paint swatch (Figure3d). We contribute printable STL palettes of 25 textures wefound in an examination of this design space.

INTERNAL DESIGN TOOLPrinting models as completely solid objects is prohibitivelycostly in both material and time. In additive manufacturingprocesses such as Fused Filament Fabrication (FFF)7, slicingsoftware converts a 3D model file into correctly sized lay-ers and toolpaths for printing. In the slicing stage, modelsare commonly hollowed and filled with a structural supportcalled infill. Our internal design tool uses the open-sourceUltimaker CuraEngine8 - itself a slicer - to generate customg-code toolpaths. We exploit the infill to control the kinaes-thetic characteristics of an object and develop an interface fordesigning these elements into 3D models.

Internal fill patternsFive infill patterns were used to control the compliance alongeach dimension of the 3D model: line, grid, concentric,grided spiral, and weighted chamber. The first three patternsare provided by the CuraEngine project. Figure 5 shows eachinfill pattern and the corresponding directions of compliance.The concentric pattern, for example, is unsupported in X andY but stiff in the Z direction. Since infills typically have thin-ner walls and are printed at faster speeds, patterns typicallyare homogenous along the z-dimension. For the grided spi-ral, we achieved compliance in 3 dimensions by rotating thegrid infill with each layer at a rate of 0.17 rad/mm. The weightchamber infill (Figure 4B) was designed to provide both sup-port but also a permeable structure. This is used so that a usermight fill the model post-print with a material and prototypethe weight characteristics in situ. The fillable portion, or dis-tribution, of the weight infill can be specified to be: offset,centralized, or completely fillable.7Fused Filament Fabrication is synonymous with Fused DepositionModeling (FDM), a process patented by Stratasys Ltd.8https://github.com/Ultimaker/CuraEngine

4

weight chamber fill

fill with a heavier material

A CB weight chamber fillB D weighted base

Figure 4. The internal design tool. (a) A model is divided into chambers. Users can specify what type of infill each chamber should have. (b) A crosssection of the bottom weight chamber of the bunny; filled with black sand for contrast. (c) The user manually injects the infill with a heavier material.(d) The final bunny artifact weighted down by green acrylic medium.

Selecting infillOur internal tool first loads a user-provided STL modelfile, identifies points of interaction, and specifies the desiredprint direction. This is then used to split the model into z-chambers, which a user can then use to specify infill prop-erties (Figure 4A). Users have control of the fill pattern, filldensity, and in the case of the weight chamber, distribution.

KinaesthesiaIn our internal design tool, a model is separated into equalsized chambers (Figure 4A). A user can alter the chamberboundaries to correspond with semantic boundaries (e.g. thetail of a bunny). A user can then select an appropriate hap-tic infill that is parametrized through infill density. Though apercentage fill is common in existing slicers, this allows ourinfill patterns to change stiffness. We printed each infill in arange of percentages and then tested their compliance on acustom stress-strain fixture (Figure 9). We then mapped eachinfill percentage to the measured compliance. Thus a usercan specify an estimated stiffness and direction(s) of flexi-bility which is then used by the internal tool to generate theappropriate infill and density.

To achieve a desired weight and mass distribution, materi-als can be injected into objects post-print using the weightedchamber infill. This infill functions in its primary capacityi.e. supporting outer walls during print-time, but also leaves arelatively hollow chamber that allows fluid materials to fill achamber of the model through a small pipe. This is achievedby routing the liquid through channels in between supportwalls (Figure 4C). This provides unique opportunities for theuser to alter the balance and kinesthetic properties of the ob-ject. A user might for instance wish to create a stabilizingbase. The bottom chamber of the model can be filled with a

line grid concentric doubleconcentric

weight weightaxis full

Figure 5. Various patterns of infill provide different types of compliancealong dimensions orthogonal to the build direction. Red arrows indicateresistive (less compliant) points whereas blue arrows designate compli-ant directions. The weighted infill is designed such that a medium is ableto be injected post print.

heavier material (Figure 4D). An artifact may have multipleweight chambers or other infill chambers; to prevent interfer-ence between chambers, a 0.8 mm solid chamber wall existsaround chamber boundaries.

FABRICATION TECHNIQUE AND EVALUATIONIn this section we describe our printing techniques, evalua-tion, as well as example physical artifacts produced by eachtool. We discuss the potential and limitations of each method.

Tools and MaterialsFor fabrication, we used a Type A Series 1 Machine (TAM)with a G2, 0.4 mm extruder. Two types of filament were used:1.75 mm PLA, a strong and durable plastic, and 1.75 mm Nin-jaFlex TPE, a flexible thermoplastic elastomer. All artifactswere printed at maximum resolution (0.10 mm layer height).TAMs share a common architecture with most FFF machines;it extrudes thermoplastic filament through a heated nozzle andmotion is derived from a 3-axis gantry. Stereolithography(SLA) and PolyJet printing differs from FFF by jetting or cur-ing polymer resin in successive layers to build up a model; al-though not explicitly tested our approach is extendable sinceit modifies common procedures (i.e. support and infill mate-rial) typical of most additive manufacturing process.

Tactile EvaluationTo evaluate our external design tool, we printed a diverse setof reference objects and evaluated print quality. A chief con-cern of the FFF process is that print quality is highly depen-dent on planar orientation. A grain in Fused Filament Fabri-cation, for instance, results from small ridges between layers.

The printed set consisted of: a) the library of 25 tactile sur-faces in PLA on 70 mm x 70 mm squares with a 2 mm base,and b) three representative textures printed on cones, spheres,planes, and cylinders primitives, shown in Figure 6. Theseprimitives were chosen as to cover all 3D planar orienta-tions. Furthermore, these primitives explore how these tex-tures feel on various common objects through different ex-ploratory haptic interactions (e.g., pinch, stroke, enclosure).All features were displaced less than 3 mm from the surfacenormal.

We found that printing objects with textures had the practi-cal benefit of masking the grain of 3D printed artifacts. Thistechnique was limited by retraction errors, where extrudedmaterial was not properly retracted and resulted in cobwebtraversals (e.g. traveling between peaks in spiky textures).

5

Figure 6. 3 cylinders and 3 cones printed with alligator (indexical),bump (iconic), and arrow (symbolic) textures, respectively.

This could easily be improved with some post-print cleaning.As perceived by the authors, small features sizes (< 0.5 mm)and feature gaps (< 0.1 mm) produced smoother surfaces.Print times for these parts increased from 5-15%, dependingon feature-area. Textures with smaller features than the layerresolution (< 0.1 mm) - like knurl, or alligator - printed bestin all planes and obscured the artifact grain. However thesetextures were less easily identifiable. Textures with larger fea-tures - like corn or arrows - did not hide the layering but weremore easily identifiable. In future work we will investigatelayering large features with a smoothing noise texture to pro-duce both visually and tactually pleasing objects.

Kinaesthetics EvaluationOur goal in printing and evaluating the various infill pat-terns was to understand and control how each pattern var-ied along each axes for artifacts printed with flexible filament(TPE). Since TPE agglomerates at narrow points like conesand spheres (Figure 6), each infill pattern was printed in 30mm cubes.

More formally, the compliance of a material is its elastic mod-ulus, where the material deforms under stress yet returns tooriginal position9. We modeled each object as a linear springand report the spring rate as ∆F

∆y in units of Nm . The mea-

surement system was built with a linear displacement gauge,a digital scale, a drill press, and a custom bracket to con-nect the components; the setup is shown in Figure 8. Thedrill press applied load to the sample piece while the distancegauge measured deflection and the scale measured load.

We show the compliance testing for an infill pattern that iscompliant in two-axes - results shown in Figure 9. We printedfour cubes varying in fill density from 20% to 80%, took>5 data points, and found the linear regression with the y-intercept set to y = 0. In all cases the r2 > 94% when de-pressed by 50%, meaning we can reliably predict deformation9This is the linear region of the stress-strain plot. Plastic deforma-tion is permanent strain to a material. We only measured the com-pression since most UI elements are rarely in tension.

dial indicatorprinted fixture

printed flexible object

digital scale

drill press

Figure 8. Our compliance testing setup measured force with respect toapplied displacement. The drill press was manually operated.

to half an object’s height. We also found that 80% and 100%infills were effectively rigid, so flexibility and rigidity canbe printed into a single part. For reference, overall pinchingstrength varies from 50 to 100 N for average adults [20], thusa 20% infill would feel completely deformable while 80% in-fill would feel nearly rigid. In our design tool, this measurewas linearized for user-selection.

Figure 9. Linear regressions for various percentages of infill and theircorresponding compliance. Spring rates require a y-intercept at 0.

For weighted artifacts, we injected a variety of materials post-print including: hot glue, epoxies, sands, and acrylic medium.A pipette was used to pipe liquid mediums into appropriateweight chambers, shown in Figure 4C. Hot glue was used toclose the resulting hole. Epoxies were allowed to cure beforeclosure. Due to imperfect layer fusion from FFF, we foundthat chambers needed to have wall thicknesses of at least 0.8mm in order to contain less viscous fillers. Filling these cav-ities with conductive materials may be utilized to provide in-teractive properties to otherwise passive prints.

FEEL AESTHETICS IN EXAMPLE OBJECTSIn order to evaluate how our tool expands the current feel aes-thetic of 3D printed objects, we designed a set of four pro-totype objects using both internal and external HapticPrinttools. Each object explores how haptic properties can be uti-lized and prototyped in an iterative design process.

Affordance Design for User InterfacesThe physical affordances, or cues that convey potential ac-tion, are a powerful design property; however, capacitivetouch surfaces have few physical affordances. Harrison etal. explored pneumatically actuated latex templates to pro-vide some tactile information [9]. We created a set of threepassive UI interfaces that offer a higher resolution of hap-tic information. These UIs are used in a common interface

6

A B C D

Figure 7. (a) A fan control board. A multi-button pattern is attached to a capacitive layer (bottom) and a raised texture-enchanced surface is placedto give the interface tactile cues and affordances. (b) A weighted die. Acrylic medium is pipetted into an axial chamber. (top) The die backlit. (c) Aprintmaking wheel with texture, compliance, and weight. (d) A blade handle with additional pressure cues to indicate correct finger placement.

control problem of device controls. Notably, we created thedesigns using 2D graphics and SVG editors.

To showcase how haptic design can be used to enhance inter-active artifacts, we built a capacitive touch sensing layer us-ing an MPR121. A TPE texture layer (dielectric) was placedover copper tape (capactive layer) to give tactile cues to theuser (Figure 10). The dielectric was fabricated at a 15% infill,which we found as a usable button stiffness. For displace-ments less than 4 mm from the copper electrode, we foundthat two or more discrete events such as “touch” and “press”could be reliably detected.

This technique was applied to a simple four-button controlboard to control an connected 3-speed fan (Figure 7a). Thebuttons are circles with radial gradients, displaced 4mm froma planar surface, and constructed with a compliant infill. Theicons were overlaid on the buttons and then textures wereclip-masked around two conceptual control areas to aid withidentification. A spiky texture was customized by lookingup feature-size and feature-gap to obtain a desired roughnessusing the HapticPrint texture palette. The same design wasthen used to vinyl cut copper traces, which were affixed to alaser-cut acrylic backing (Figure 10).

For interaction, we formatted the buttons to have two degreesof “actionability”. We utilized our capacitive displacementtechnique to turn off the fan gradually from a light touch, orimmediately with a hard press. Thus, the affordances of thebutton were used to communicate multiple physical interac-tions. We created a second interface that used striated marksto guide the finger up and down to communicate the affor-dance of slide-ability. In lighting control situations, customtactile patterns could be derived from building floor plans,providing a spatial aid.

A B C

Figure 10. a) An interface design with “spikey” button and “striated”slider elements. b) The printed design in NinjaFlex, with its correspond-ing copper traces affixed to support three discrete slider positions. c)The sliding mechanism detected touch which triggered an LED.

Weight as functionWeight is regularly used to create stable configurations for ob-jects, however it can also be leveraged to provide additionalfunctionality to objects. Figure 7b depicts a weighted die thathas been applied an axial weight pattern. This biases weight(and more literally a craps game) to favor rotary interactions.We then utilized these haptic properties together in a print-making tool that maps a 2D texture of a tire pattern arounda cylinder. We applied a texture to the wheel, then applied acompliant infill to the exterior of the wheel with a weightedaxial chamber. We injected the wheel with an acrylic mediumfor a more realistic moment-of-inertia. Figure 7c shows thefinal product.

Handle for Digital ApprenticeshipIn a final application, we looked at how custom tools couldbe modified to have added usability. We added compliantsurfaces to areas of a blade handle that corresponds with areaswhere pressure was being placed by an expert user (Figure7d). In this way, a user who downloaded and printed thishandle could learn tacit knowledge from subtle pressure cues.A capacitive press sensor could likewise provide feedback tothe user to provide feedback or act as a kill switch to preventuser injury if held improperly.

DISCUSSIONFeel aesthetics are critical to the design of everyday objects,but this space is not captured by design tools or easily embed-ded within the tooling constraints of the fabrication process.Tools like HapticPrint can lead to designs that are more us-able (affordances), accessible (improved support for peoplewith disabilities), and aesthetically diverse (beyond smoothand uniform surfaces). However key challenges exist to inte-grating physical characteristics into virtual design tools.

Haptic design is inherently a tangible, physical design pro-cess. This raises the challenge of integrating haptics into avirtual design environment. HapticPrint addressed this issueby hand-curating a self-referential palette of textures, infills,and weight distributions, reminiscent of current practice withcolor swatches or color systems. This system however is pref-aced on atomic haptic units, whereas real-world objects are amixtures and layers of tactual and kinaesthetic feels. Dither-ing the infill, for instance, could create different haptic pro-files. Selection tasks could be mitigated by generating small

7

multiples and purposefully generating principled feature rep-resentations for data-driven design [30].

For end-user-created feel aesthetics, additional feedbackcould be provided by performing a similarity search over acorpus of existing textures or developing a higher-level modelof expected behavior. HapticPrint found reasonable feel char-acterizations along feature size, gap, and density; howevera formal psychophysics user study is needed to verify thismechanism. Our post-weight print technique provides flexi-bility for the user to choose the appropriate weight distribu-tion in situ for a more reflective design process.

Alternatively, feel can be part of a goal-centered design pro-cess where known material behaviors are matched to desiredbehaviors [1], however the haptic vocabulary even amongstprofessional designers is limited [29]. Fully communicatingmultifaceted “feel”’ remains a subject for future work.

In the context of industrial-scale fabrication, additive manu-facturing (AM) is primarily limited by speed and resolution.Rather than compete with injection molding on speed, AMexcels in “mass customization”, reducing the cost for one-offdesigns and trade-offs for added complexity. Though micro-textures are common in injection molding (plastic) or elec-troforming (metal), little work exists for these types of fin-ishing textures (e.g. matte, satin, gloss) in printed parts. Re-cent work has demonstrated the feasibility of controlling mi-crogeometries to fabricate custom reflectance properties [34].HapticPrint may apply to custom manufacturing by control-ling microtextures without the complex mold-making pro-cesses. Additionally, we achieved textures at the resolutionof most printing technologies that further helped mask FFFgrain leading to more appealing final products. Although ourcurrent fill method for augmenting weight is labor-intensive,we see this as a near term solution that will be mediated inthe future by multi-material printing.

As multi-material fabrications tools advance, the expandedcollection of material properties can be leveraged both struc-turally (density) and dynamically (viscosity) and lead tounique haptic experiences. HapticPrint contributes a first looktowards how these properties might be expressed through avirtual design tool. By enabling haptic design to be moreeasily designed and produced within 3D designs, we have en-abled a broader design space for experts and novices to ex-plore. We are hopeful that this approach will further designtools towards becoming medium-aware and engaging withphysical design processes.

FUTURE WORK & LIMITATIONSWe evaluated fabrication techniques for expanding the hapticvocabulary of 3D prints. However in order to fully evalu-ate the haptic characteristics of artifacts, our evaluation waslimited by a lack of a formal psychophysics study. Deter-mining elementary measures such as perceptual thresholds,or derived measures such as hardness and smoothness, couldbetter describe the haptic space achievable (and unachievable)using our technique. Alternative approaches include a Gibso-nian view of the “haptic system” which conversely eschewsatomic characterization of touch and kinesthetics for a more

holistic evaluation [6]. This would be a more apt methodol-ogy for haptic design since human observers are free to ex-plore, move, or change patterns of touch; this would result ina more subjective report of perception, as would be the casewhen composing multiple haptic sensations in a design.

This paper addresses a larger breadth of Klatzky’s haptic ex-ploration than the state-of-the-art in digital fabrication. Hap-ticPrint provides additional creative handles for controllinghaptic information through pressure, contour following andlateral motion, and unsupported holding. We see great valuefrom gaining haptic information from static contact, mostcommonly achieved through thermal stimulation, which canprovide additional cues to users and opportunities for inter-action design (e.g. object identification [13]). More interac-tive haptic cues can be derived through a temporal dimension(such as time-varying textures [7]), or from chemical, electri-cal, or mechanical stimulation [31].

CONCLUSIONFeel Aesthetics are often lacking in printed designs. In thispaper, we identified methods and built tools for easily incor-porating feel aesthetics into 3D designs. Our HapticPrint in-terfaces allow a user to easily modify the internal and externalhaptics through 2D design, which is a medium more widelyunderstood. We evaluated HapticPrint by printing referenceartifacts spanning the haptic space and assessed print quality.In a set of example objects, we showcased how tactility andkinaesthetics could be used to design more expressive, usable,and accessible 3D artifacts.

ACKNOWLEDGMENTSWe thank anonymous reviewers for their insightful com-ments. This research was supported by NSF Grant No. IIS-1451465 and a NSF Graduate Research Fellowship.

REFERENCES1. Bickel, B., Bcher, M., Otaduy, M. A., Lee, H. R., Pfister,

H., Gross, M., and Matusik, W. Design and fabricationof materials with desired deformation behavior. Trans.on Graphics 29, 4 (July 2010), 1.

2. Brown, C., and Hurst, A. VizTouch: Automaticallygenerated tactile visualizations of coordinate spaces. InProc. of TEI’12, ACM (2012), 131–138.

3. Clark, F. J., and Horch, K. W. Kinesthesia. Handbook ofPerception and Human Performance 1 (1986).

4. Echevarria, J. I., Bradley, D., Gutierrez, D., and Beeler,T. Capturing and stylizing hair for 3D fabrication. Trans.on Graphics 33, 4 (July 2014), 1–11.

5. Furferi, R., Governi, L., Volpe, Y., Puggelli, L., Vanni,N., and Carfagni, M. From 2D to 2.5D i.e. from paintingto tactile model. Graphical Models 76 (2014), 706–723.

6. Gibson, J. J. The Senses Considered as PerceptualSystems. Houghton Mifflin, 1966.

7. Gu, J., Tu, C.-I., Ramamoorthi, R., Belhumeur, P.,Matusik, W., and Nayar, S. Time-varying surfaceappearance: Acquisition, modeling and rendering. InACM Trans. on Graphics, vol. 25 (2006), 762–771.

8

8. Gtzelmann, T., and Pavkovic, A. Towards automaticallygenerated tactile detail maps by 3D printers for blindpersons. In Computers Helping People with SpecialNeeds. Springer, 2014, 1–7.

9. Harrison, C., and Hudson, S. E. Providing dynamicallychangeable physical buttons on a visual display. In InProc. of the CHI’09 Conference on Human Factors inComputing Systems, ACM (2009), 299–308.

10. Hollins, M., Faldowski, R., Rao, S., and Young, F.Perceptual dimensions of tactile surface texture: Amultidimensional scaling analysis. Perception &Psychophysics 54, 6 (1993), 697–705.

11. Houde, S. What Do Prototypes Prototype? In Handbookof Human Computer Interaction, 2 ed. Elsevier Science,1997.

12. Hudson, S. E. Printing teddy bears: A technique for 3Dprinting of soft interactive objects. In Proc. of UIST’14,ACM Press (2014), 459–468.

13. Jones, L., Berris, M., and others. The psychophysics oftemperature perception and thermal-interface design. InIn Proc. of HAPTICS’02, IEEE (2002), 137–142.

14. Klatzky, R. L., and Lederman, S. J. Touch. John Wiley &Sons, Inc., 2003.

15. Kreifeldt, J., Lin, R., and Chuang, M.-C. Theimportance of feel in product design feel, the neglectedaesthetic Do Not Touch. In Internationalization, Designand Global Development. Springer, 2011, 312–321.

16. Lederman, S., and Klatzky, R. Hand Movements: AWindow into Haptic Object Recognition. Cognitivepsychology 19 (1987), 342–368.

17. Li, H., Vouga, E., Gudym, A., Luo, L., Barron, J. T., andGusev, G. 3D self-portraits. Trans. on Graphics 32, 6(Nov. 2013), 1–9.

18. Loomis, J. M., and Lederman, S. J. Tactual perception.Handbook of Perception and Human Performances 2(1986).

19. Lu, L., Chen, B., Sharf, A., Zhao, H., Wei, Y., Fan, Q.,Chen, X., Savoye, Y., Tu, C., and Cohen-Or, D.Build-to-last: Strength to weight 3D printed objects.Trans. on Graphics 33, 4 (July 2014), 1–10.

20. Mathiowetz, V., Kashman, N., Volland, G., Weber, K.,Dowe, M., Rogers, S., and others. Grip and pinchstrength: Normative data for adults. 69–74.

21. Mueller, S., Im, S., Gurevich, S., Teibrich, A., Pfisterer,L., Guimbretire, F., and Baudisch, P. WirePrint: 3Dprinted previews for fast prototyping. In Proc. ofUIST’14, ACM Press (2014), 273–280.

22. Norman, D. The Psychology of Everyday Things. BasicBooks, 1998.

23. Prevost, R., Whiting, E., Lefebvre, S., andSorkine-Hornung, O. Make it stand: Balancing shapesfor 3D fabrication. Trans. on Graphics 32 (2013).

24. Reichinger, A., Neumller, M., Rist, F., Maierhofer, S.,and Purgathofer, W. Computer-Aided Design of TactileModels. Springer, 2012.

25. Savage, V., Schmidt, R., Grossman, T., Fitzmaurice, G.,and Hartmann, B. A series of tubes: Adding interactivityto 3D prints using internal pipes. In Proc. of UIST’14,ACM Press (2014), 3–12.

26. Srinivasan, M. A., Whitehouse, J. M., and LaMotte,R. H. Tactile detection of slip: Surface microgeometryand peripheral neural codes. 1323–1332.

27. Stava, O., Vanek, J., Benes, B., Carr, N., and Mch, R.Stress relief: Improving structural strength of 3Dprintable objects. Trans. on Graphics 31, 4 (2012), 1–11.

28. Swaminathan, S., Shi, C., Jansen, Y., Dragicevic, P.,Oehlberg, L. A., and Fekete, J.-D. Supporting the designand fabrication of physical visualizations. ACM Press(2014), 3845–3854.

29. Teinaki, V., Montgomery, B., Spencer, N., and Cockton,G. An aesthetics of touch: Investigating the language ofdesign relating to form.

30. Torres, C., and Paulos, E. MetaMorphe: Designingexpressive 3D models for digital fabrication. In Proc. ofC&C’15, 73–82.

31. Van Erp, J. B., Kyung, K.-U., Kassner, S., Carter, J.,Brewster, S., Weber, G., and Andrew, I. Setting thestandards for haptic and tactile interactions: ISOs work.In Haptics: Generating and Perceiving TangibleSensations. Springer, 353–358.

32. Vidime, K., Wang, S.-P., Ragan-Kelley, J., and Matusik,W. OpenFab: A programmable pipeline formulti-material fabrication. Trans. on Graphics 32, 4(July 2013), 1.

33. Wang, W., Wang, T. Y., Yang, Z., Liu, L., Tong, X.,Tong, W., Deng, J., Chen, F., and Liu, X. Cost-effectiveprinting of 3D objects with skin-frame structures. Trans.on Graphics 32, 6 (Nov. 2013), 1–10.

34. Weyrich, T., Peers, P., Matusik, W., and Rusinkiewicz,S. Fabricating microgeometry for custom surfacereflectance. In Trans. on Graphics (TOG), vol. 28, 32.

35. Willis, K., Brockmeyer, E., Hudson, S., and Poupyrev, I.Printed optics: 3D printing of embedded opticalelements for interactive devices. In Proc. of UIST’12,ACM (2012), 589–598.

36. Willis, K. D. D., and Wilson, A. D. InfraStructs:Fabricating information inside physical objects forimaging in the terahertz region. 138:1–138:10.

37. Yao, L., Niiyama, R., Ou, J., Follmer, S., Della Silva, C.,and Ishii, H. PneUI: Pneumatically actuated softcomposite materials for shape changing interfaces. ACMPress (2013), 13–22.

9